eukara/ngunix
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
ngunix/tracker/it/reads3m.c

672 lines
17 KiB
C
Raw Blame History

/* _______ ____ __ ___ ___
* \ _ \ \ / \ / \ \ / / ' ' '
* | | \ \ | | || | \/ | . .
* | | | | | | || ||\ /| |
* | | | | | | || || \/ | | ' ' '
* | | | | | | || || | | . .
* | |_/ / \ \__// || | |
* /_______/ynamic \____/niversal /__\ /____\usic /| . . ibliotheque
* / \
* / . \
* reads3m.c - Code to read a ScreamTracker 3 / / \ \
* module from an open file. | < / \_
* | \/ /\ /
* By entheh. \_ / > /
* | \ / /
* | ' /
* \__/
*/
#ifdef DUMB
// IT_STEREO... :o
#include <stdlib.h>
#include <string.h>
#include "dumb.h"
#include "internal/it.h"
/** WARNING: this is duplicated in itread.c */
static int it_seek(DUMBFILE *f, long offset)
{
long pos = dumbfile_pos(f);
if (pos > offset)
return -1;
if (pos < offset)
if (dumbfile_skip(f, offset - pos))
return -1;
return 0;
}
static int it_s3m_read_sample_header(IT_SAMPLE *sample, long *offset, DUMBFILE *f)
{
unsigned char type;
int flags;
type = dumbfile_getc(f);
if (type > 1) {
/** WARNING: no adlib support */
}
dumbfile_getnc(sample->filename, 13, f);
sample->filename[13] = 0;
*offset = dumbfile_igetw(f) << 4;
sample->length = dumbfile_igetl(f);
sample->loop_start = dumbfile_igetl(f);
sample->loop_end = dumbfile_igetl(f);
sample->default_volume = dumbfile_getc(f);
dumbfile_skip(f, 1);
if (dumbfile_getc(f) != 0)
/* Sample is packed apparently (or error reading from file). We don't
* know how to read packed samples.
*/
return -1;
flags = dumbfile_getc(f);
sample->C5_speed = dumbfile_igetl(f) << 1;
/* Skip four unused bytes and three internal variables. */
dumbfile_skip(f, 4+2+2+4);
dumbfile_getnc(sample->name, 28, f);
sample->name[28] = 0;
if (type == 0) {
/* Looks like no-existy. Anyway, there's for sure no 'SCRS' ... */
sample->flags &= ~IT_SAMPLE_EXISTS;
return dumbfile_error(f);
}
if (dumbfile_mgetl(f) != DUMB_ID('S','C','R','S'))
return -1;
sample->global_volume = 64;
sample->flags = IT_SAMPLE_EXISTS;
if (flags & 1) sample->flags |= IT_SAMPLE_LOOP;
if (flags & 2) sample->flags |= IT_SAMPLE_STEREO;
if (flags & 4) sample->flags |= IT_SAMPLE_16BIT;
sample->default_pan = 0; // 0 = don't use, or 160 = centre?
if (sample->length <= 0)
sample->flags &= ~IT_SAMPLE_EXISTS;
else if (sample->flags & IT_SAMPLE_LOOP) {
if ((unsigned int)sample->loop_end > (unsigned int)sample->length)
sample->flags &= ~IT_SAMPLE_LOOP;
else if ((unsigned int)sample->loop_start >= (unsigned int)sample->loop_end)
sample->flags &= ~IT_SAMPLE_LOOP;
else
/* ScreamTracker seems not to save what comes after the loop end
* point, but rather to assume it is a duplicate of what comes at
* the loop start point. I am not completely sure of this though.
* It is easy to evade; simply truncate the sample.
*/
sample->length = sample->loop_end;
}
//Do we need to set all these?
sample->vibrato_speed = 0;
sample->vibrato_depth = 0;
sample->vibrato_rate = 0;
sample->vibrato_waveform = IT_VIBRATO_SINE;
return dumbfile_error(f);
}
static int it_s3m_read_sample_data(IT_SAMPLE *sample, int ffi, DUMBFILE *f)
{
long n;
long datasize = sample->length;
if (sample->flags & IT_SAMPLE_STEREO) datasize <<= 1;
sample->data = malloc(datasize * (sample->flags & IT_SAMPLE_16BIT ? 2 : 1));
if (!sample->data)
return -1;
if (sample->flags & IT_SAMPLE_STEREO) {
if (sample->flags & IT_SAMPLE_16BIT) {
for (n = 0; n < datasize; n += 2)
((short *)sample->data)[n] = dumbfile_igetw(f);
for (n = 1; n < datasize; n += 2)
((short *)sample->data)[n] = dumbfile_igetw(f);
} else {
for (n = 0; n < datasize; n += 2)
((signed char *)sample->data)[n] = dumbfile_getc(f);
for (n = 1; n < datasize; n += 2)
((signed char *)sample->data)[n] = dumbfile_getc(f);
}
} else if (sample->flags & IT_SAMPLE_16BIT)
for (n = 0; n < sample->length; n++)
((short *)sample->data)[n] = dumbfile_igetw(f);
else
for (n = 0; n < sample->length; n++)
((signed char *)sample->data)[n] = dumbfile_getc(f);
if (dumbfile_error(f))
return -1;
if (ffi != 1) {
/* Convert to signed. */
if (sample->flags & IT_SAMPLE_16BIT)
for (n = 0; n < datasize; n++)
((short *)sample->data)[n] ^= 0x8000;
else
for (n = 0; n < datasize; n++)
((signed char *)sample->data)[n] ^= 0x80;
}
return 0;
}
static int it_s3m_read_pattern(IT_PATTERN *pattern, DUMBFILE *f, unsigned char *buffer)
{
int buflen = 0;
int bufpos = 0;
IT_ENTRY *entry;
unsigned char channel;
/* Haha, this is hilarious!
*
* Well, after some experimentation, it seems that different S3M writers
* define the format in different ways. The S3M docs say that the first
* two bytes hold the "length of [the] packed pattern", and the packed
* pattern data follow. Judging by the contents of ARMANI.S3M, packaged
* with ScreamTracker itself, the measure of length _includes_ the two
* bytes used to store the length; in other words, we should read
* (length - 2) more bytes. However, aryx.s3m, packaged with ModPlug
* Tracker, excludes these two bytes, so (length) more bytes must be
* read.
*
* Call me crazy, but I just find it insanely funny that the format was
* misunderstood in this way :D
*
* Now we can't just risk reading two extra bytes, because then we
* overshoot, and DUMBFILEs don't support backward seeking (for a good
* reason). Luckily, there is a way. We can read the data little by
* little, and stop when we have 64 rows in memory. Provided we protect
* against buffer overflow, this method should work with all sensibly
* written S3M files. If you find one for which it does not work, please
* let me know at entheh@users.sf.net so I can look at it.
*/
/* Discard the length. */
dumbfile_skip(f, 2);
if (dumbfile_error(f))
return -1;
pattern->n_rows = 0;
pattern->n_entries = 0;
/* Read in the pattern data, little by little, and work out how many
* entries we need room for. Sorry, but this is just so funny...
*/
for (;;) {
unsigned char b = buffer[buflen++] = dumbfile_getc(f);
#if 1
static const unsigned char used[8] = {0, 2, 1, 3, 2, 4, 3, 5};
channel = b & 31;
b >>= 5;
pattern->n_entries++;
if (b) {
if (buflen + used[b] >= 65536) return -1;
dumbfile_getnc(buffer + buflen, used[b], f);
buflen += used[b];
} else {
/* End of row */
if (++pattern->n_rows == 64) break;
if (buflen >= 65536) return -1;
}
#else
if (b == 0) {
/* End of row */
pattern->n_entries++;
if (++pattern->n_rows == 64) break;
if (buflen >= 65536) return -1;
} else {
static const unsigned char used[8] = {0, 2, 1, 3, 2, 4, 3, 5};
channel = b & 31;
b >>= 5;
if (b) {
pattern->n_entries++;
if (buflen + used[b] >= 65536) return -1;
dumbfile_getnc(buffer + buflen, used[b], f);
buflen += used[b];
}
}
#endif
/* We have ensured that buflen < 65536 at this point, so it is safe
* to iterate and read at least one more byte without checking.
* However, now would be a good time to check for errors reading from
* the file.
*/
if (dumbfile_error(f))
return -1;
}
pattern->entry = malloc(pattern->n_entries * sizeof(*pattern->entry));
if (!pattern->entry)
return -1;
entry = pattern->entry;
while (bufpos < buflen) {
unsigned char b = buffer[bufpos++];
#if 1
if (!(b & ~31))
#else
if (b == 0)
#endif
{
/* End of row */
IT_SET_END_ROW(entry);
entry++;
continue;
}
channel = b & 31;
if (b & 224) {
entry->mask = 0;
entry->channel = channel;
if (b & 32) {
unsigned char n = buffer[bufpos++];
if (n != 255) {
if (n == 254)
entry->note = IT_NOTE_CUT;
else
entry->note = (n >> 4) * 12 + (n & 15);
entry->mask |= IT_ENTRY_NOTE;
}
entry->instrument = buffer[bufpos++];
if (entry->instrument)
entry->mask |= IT_ENTRY_INSTRUMENT;
}
if (b & 64) {
entry->volpan = buffer[bufpos++];
if (entry->volpan != 255)
entry->mask |= IT_ENTRY_VOLPAN;
}
if (b & 128) {
entry->effect = buffer[bufpos++];
entry->effectvalue = buffer[bufpos++];
if (entry->effect != 255) {
entry->mask |= IT_ENTRY_EFFECT;
if (entry->effect == IT_BREAK_TO_ROW)
entry->effectvalue -= (entry->effectvalue >> 4) * 6;
}
/** WARNING: ARGH! CONVERT TEH EFFECTS!@~ */
}
entry++;
}
}
ASSERT(entry == pattern->entry + pattern->n_entries);
return 0;
}
/** WARNING: this is duplicated in itread.c - also bad practice to use the same struct name unless they are unified in a header */
/* Currently we assume the sample data are stored after the sample headers in
* module files. This assumption may be unjustified; let me know if you have
* trouble.
*/
#define IT_COMPONENT_INSTRUMENT 1
#define IT_COMPONENT_PATTERN 2
#define IT_COMPONENT_SAMPLE 3
typedef struct IT_COMPONENT
{
unsigned char type;
unsigned char n;
long offset;
short sampfirst; /* component[sampfirst] = first sample data after this */
short sampnext; /* sampnext is used to create linked lists of sample data */
}
IT_COMPONENT;
static int it_component_compare(const void *e1, const void *e2)
{
return ((const IT_COMPONENT *)e1)->offset -
((const IT_COMPONENT *)e2)->offset;
}
static DUMB_IT_SIGDATA *it_s3m_load_sigdata(DUMBFILE *f)
{
DUMB_IT_SIGDATA *sigdata;
int flags, cwtv, ffi;
int default_pan_present;
IT_COMPONENT *component;
int n_components = 0;
int n;
unsigned char *buffer;
sigdata = malloc(sizeof(*sigdata));
if (!sigdata) return NULL;
dumbfile_getnc(sigdata->name, 28, f);
sigdata->name[28] = 0;
if (dumbfile_getc(f) != 0x1A || dumbfile_getc(f) != 16) {
free(sigdata);
return NULL;
}
dumbfile_skip(f, 2);
sigdata->song_message = NULL;
sigdata->order = NULL;
sigdata->instrument = NULL;
sigdata->sample = NULL;
sigdata->pattern = NULL;
sigdata->midi = NULL;
sigdata->checkpoint = NULL;
sigdata->n_orders = dumbfile_igetw(f);
sigdata->n_instruments = 0;
sigdata->n_samples = dumbfile_igetw(f);
sigdata->n_patterns = dumbfile_igetw(f);
if (dumbfile_error(f) || sigdata->n_orders <= 0 || sigdata->n_samples > 256 || sigdata->n_patterns > 256) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
sigdata->order = malloc(sigdata->n_orders);
if (!sigdata->order) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
if (sigdata->n_samples) {
sigdata->sample = malloc(sigdata->n_samples * sizeof(*sigdata->sample));
if (!sigdata->sample) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
for (n = 0; n < sigdata->n_samples; n++)
sigdata->sample[n].data = NULL;
}
if (sigdata->n_patterns) {
sigdata->pattern = malloc(sigdata->n_patterns * sizeof(*sigdata->pattern));
if (!sigdata->pattern) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
for (n = 0; n < sigdata->n_patterns; n++)
sigdata->pattern[n].entry = NULL;
}
flags = dumbfile_igetw(f);
cwtv = dumbfile_igetw(f);
if (cwtv == 0x1300) {
/** WARNING: volume slides on every frame */
}
ffi = dumbfile_igetw(f);
/** WARNING: which ones? */
sigdata->flags = IT_STEREO | IT_OLD_EFFECTS | IT_COMPATIBLE_GXX;
if (dumbfile_mgetl(f) != DUMB_ID('S','C','R','M')) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
sigdata->global_volume = dumbfile_getc(f) << 1;
sigdata->speed = dumbfile_getc(f);
if (sigdata->speed == 0) sigdata->speed = 6; // Should we? What about tempo?
sigdata->tempo = dumbfile_getc(f);
/*master_volume = */dumbfile_getc(f); // 7 bits; +128 for stereo
//what do we do with master_volume? it's not the same as mixing volume...
sigdata->mixing_volume = 48;
/* Skip GUS Ultra Click Removal byte. */
dumbfile_getc(f);
default_pan_present = dumbfile_getc(f);
dumbfile_skip(f, 8);
/* Skip Special Custom Data Pointer. */
/** WARNING: investigate this? */
dumbfile_igetw(f);
/* Channel settings for 32 channels, 255=unused, +128=disabled */
{
int i;
for (i = 0; i < 32; i++) {
int c = dumbfile_getc(f);
if (!(c & (128 | 16))) { /* +128=disabled, +16=Adlib */
sigdata->channel_volume[i] = 64;
sigdata->channel_pan[i] = c & 8 ? 12 : 3;
/** WARNING: ah, but it should be 7 for mono... */
} else {
/** WARNING: this could be improved if we support channel muting... */
sigdata->channel_volume[i] = 0;
sigdata->channel_pan[i] = 7;
}
}
}
/* Orders, byte each, length = sigdata->n_orders (should be even) */
dumbfile_getnc(sigdata->order, sigdata->n_orders, f);
sigdata->restart_position = 0;
component = malloc(768*sizeof(*component));
if (!component) {
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
for (n = 0; n < sigdata->n_samples; n++) {
component[n_components].type = IT_COMPONENT_SAMPLE;
component[n_components].n = n;
component[n_components].offset = dumbfile_igetw(f) << 4;
component[n_components].sampfirst = -1;
n_components++;
}
for (n = 0; n < sigdata->n_patterns; n++) {
long offset = dumbfile_igetw(f) << 4;
if (offset) {
component[n_components].type = IT_COMPONENT_PATTERN;
component[n_components].n = n;
component[n_components].offset = offset;
component[n_components].sampfirst = -1;
n_components++;
} else {
/** WARNING: Empty 64-row pattern ... ? (this does happen!) */
sigdata->pattern[n].n_rows = 64;
sigdata->pattern[n].n_entries = 0;
}
}
qsort(component, n_components, sizeof(IT_COMPONENT), &it_component_compare);
/* I found a really dumb S3M file that claimed to contain default pan
* data but didn't contain any. Programs would load it by reading part of
* the first instrument header, assuming the data to be default pan
* positions, and then rereading the instrument module. We cannot do this
* without obfuscating the file input model, so we insert an extra check
* here that we won't overrun the start of the first component.
*/
if (default_pan_present == 252 && component[0].offset >= dumbfile_pos(f) + 32) {
/* Channel default pan positions */
int i;
for (i = 0; i < 32; i++) {
int c = dumbfile_getc(f);
if (c & 32)
sigdata->channel_pan[i] = c & 15;
}
}
{
int i;
for (i = 0; i < 32; i++) {
sigdata->channel_pan[i] -= (sigdata->channel_pan[i] & 8) >> 3;
sigdata->channel_pan[i] = ((int)sigdata->channel_pan[i] << 5) / 7;
}
}
sigdata->pan_separation = 128;
if (dumbfile_error(f)) {
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
buffer = malloc(65536);
if (!buffer) {
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
for (n = 0; n < n_components; n++) {
long offset;
int m;
if (it_seek(f, component[n].offset)) {
free(buffer);
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
switch (component[n].type) {
case IT_COMPONENT_PATTERN:
if (it_s3m_read_pattern(&sigdata->pattern[component[n].n], f, buffer)) {
free(buffer);
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
break;
case IT_COMPONENT_SAMPLE:
if (it_s3m_read_sample_header(&sigdata->sample[component[n].n], &offset, f)) {
free(buffer);
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
if (sigdata->sample[component[n].n].flags & IT_SAMPLE_EXISTS) {
short *sample;
for (m = n + 1; m < n_components; m++)
if (component[m].offset > offset)
break;
m--;
sample = &component[m].sampfirst;
while (*sample >= 0 && component[*sample].offset <= offset)
sample = &component[*sample].sampnext;
component[n].sampnext = *sample;
*sample = n;
component[n].offset = offset;
}
}
m = component[n].sampfirst;
while (m >= 0) {
if (it_seek(f, component[m].offset)) {
free(buffer);
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
if (it_s3m_read_sample_data(&sigdata->sample[component[m].n], ffi, f)) {
free(buffer);
free(component);
_dumb_it_unload_sigdata(sigdata);
return NULL;
}
m = component[m].sampnext;
}
}
free(buffer);
free(component);
_dumb_it_fix_invalid_orders(sigdata);
return sigdata;
}
DUH *dumb_read_s3m_quick(DUMBFILE *f)
{
sigdata_t *sigdata;
DUH_SIGTYPE_DESC *descptr = &_dumb_sigtype_it;
sigdata = it_s3m_load_sigdata(f);
if (!sigdata)
return NULL;
{
const char *tag[1][2];
tag[0][0] = "TITLE";
tag[0][1] = ((DUMB_IT_SIGDATA *)sigdata)->name;
return make_duh(-1, 1, (const char *const (*)[2])tag, 1, &descptr, &sigdata);
}
}
#endif
Reference in a new issue View git blame Copy permalink